At last month’s “SuperAI” event in Singapore, investors, consumers, and other attendees gathered around the booth of Manus, eager to know more about the startup that burst onto the artificial intelligence scene in March.
Originally based in Beijing and Wuhan, the company made waves in March when it unveiled what is described as the world’s “first general AI agent.”
An AI agent, in simple terms, is a digital assistant, a system that humans can delegate their tasks to without having to be involved in the execution.
Since its sudden rise to fame, Manus has quietly moved its headquarters to Singapore. The company has not revealed how many employees it has in the city-state, but it has started to aggressively recruit local talent this month, including AI engineers, while laying off employees in China, according to people familiar with the situation.
During a keynote speech, co-founder Zhang Tao gave his family name as Cheung, a pronunciation that shares the same Chinese character but is used by those born in Hong Kong and some born in Singapore.
And although Manus was founded by Chinese entrepreneurs, its product is in English and is not accessible in China. A senior staff member said the company has no plan to launch a Chinese version.
Users at the Singapore event had high praise for the app, saying they had not seen any comparable products on the market. Manus’s main use cases include building a website, making slides, and performing various forms of search, such as fetching real-time financial data.
RELATED ARTICLE
Since startup DeepSeek became the pride of China by singlehandedly sending Nvidia’s stock price sharply lower with its low-cost approach to AI, investor appetite for Chinese tech stocks has been revived, spurred by a fear of missing out on the homegrown AI boom in the world’s second largest economy.
That excitement has led to the search for “the next DeepSeek,” as well as interest in understanding the factors driving China’s AI advances.
While his company is focused on international markets, Cheung’s speech offered a peek into the mindset of a younger generation of Chinese tech entrepreneurs who grew up in a completely different environment than their predecessors.
Born in 1986, Cheung said he started coding when he was in primary school, using the school computers twice a week, as not a lot of families in China could afford a PC back then.
Cheung, who didn’t study overseas but previously worked for Tencent and ByteDance, said Manus spent seven months last year developing an “AI browser” with features such as automatically grouping tabs into groups by topic, and using natural language to search for people on LinkedIn.
The company ended up not releasing the product, he said, as it wasn’t user friendly enough: Human users had to keep their hands off the keyboard while AI was doing its job and controlling the browser, but it wasn’t clear when AI would be finished with its task.
Cheung’s speech underlined his company’s user-first philosophy and willingness to buck trends if necessary.
While entrepreneurs of the Jack Ma era tended to follow their then-US peers and create Chinese equivalents of American products, the younger generation is trying to innovate, not just copy and paste.
Youth and China’s increasing academic prowess are providing tailwinds for such efforts.
The DeepSeek team, for example, includes many young graduates and interns from China’s top universities, such as Tsinghua University and Peking University. State-backed laboratories and the research divisions of major US tech companies have also played a crucial role in training a large pool of local researchers.
Over half of AI researchers who earned their undergraduate degrees in China pursued graduate studies in the country and most of them worked there after graduation, according to a 2023 report by MacroPolo, the think tank of the Paulson Institute. Around a third of Chinese undergraduates went to the US for graduate school, and of those, only a tiny proportion returned to China for work, according to the same data.
“This generation of young Chinese entrepreneurs grew up during China’s economic boom, giving them access to much more information. And that is the case of DeepSeek, as its success comes from academic openness, because many AI papers have no borders,” said Tony Wang, co-founder Agora.io, a provider of real-time voice and AI-powered engagement solutions dual-based in the US and China.
Wang said he was quite surprised that DeepSeek came out of the Chinese city of Hangzhou, where he himself was born.
“The so-called elites of Silicon Valley are indeed smart, and they are driven by passion and mission. However, much of their success comes from the advantages of their environment, including top-notch venture capital, strong exit mechanisms and a concentration of global talent. But with the emergence of DeepSeek, it’s become clear that innovation can also come from other places,” he said.
By offering a robust, cost-effective open-source model, DeepSeek is enabling more individuals and businesses to adopt advanced AI solutions. This has had the knock-on effect of intensifying competition among Chinese AI firms, prompting them to enhance their own models, lower prices and consider open-sourcing their technologies.
Yefei Song, head of solutions at Ant Digital Technologies, said the company has rewritten a lot of software and offerings to the market since the “DeepSeek moment,” and that a lot of companies in China have started pursuing AI transformation rather than a more general digital transformation.
“It is a chance for everybody to grow together with this. … I think this gives a chance to change people’s mindset about China: It is not just only a follower, but also one of the leaders now in the AI era,” he said.
Song said that while China has plenty of “superapps” for almost all daily services, companies are now aiming to use AI to provide users with a more personalized experience. This entails investing in GPU servers and data centers and building their own models and agents to eventually integrate into their products.
China’s leading cloud service providers—Baidu, Alibaba and Tencent—reported 100% year-on-year growth in capital expenditures for the first quarter of the year, marking the sixth consecutive quarter where their combined investments have surpassed that of their major US counterparts.
Song expects that these domestic trends will eventually spill over into foreign markets, with more Chinese companies going abroad to build AI applications locally for local companies.
Beijing is working to support the country’s AI boom. On May 30, the State Council released the “Action Plan for Computing Power Interconnectivity,” calling for industry standards to enable all AI data centers to be linked up, and the creation of a computing power trading mechanism to ensure optimal utilization of computing resources, preventing both shortages and underuse for developers and end users alike.
James Ong, founder and managing director of the Artificial Intelligence International Institute in Singapore, said he feels the DeepSeek moment shifted the mindset within China about investing in innovation.
“I think there is an awakening among more scientists and professors to start looking into fundamental technology rather than just looking at applications,” he said. “There will be new things coming out, not just applications.”
China’s growing AI confidence comes in the face of a clampdown by the US on access to cutting-edge hardware.
Despite Washington’s restrictions on Nvidia’s H20 chips, for example, China’s appetite for AI-driven capital expenditure remains robust, according to research by Jefferies. The investment bank, citing a recent call with third-party data center service provider VNET, said China has built up sufficient chip inventories to sustain data center growth at least through the first half of 2026.
“We continue to believe China would be able to ramp up 7-[nanometer chip] foundry capacity in H2 25 and further in 2026, despite some short-term hiccup in yield likely due to adoption of local equipment,” said Jefferies in a note led by analyst Edison Lee.
Agora’s Wang was similarly upbeat about China’s prospects.
“AI will ultimately be open, and all of humanity will benefit from it. You can’t just cut it off. Even if you ban Nvidia chips from being sold to China, companies like Huawei will rise up. Maybe their performance will be 30% lower, but that’s okay if they’re three times cheaper.”
Nvidia wrote off billions of dollars in earnings due to the H20 ban, though Nikkei Asia reported in May that the US company aims to start shipping a further downgraded chip for AI applications to customers in China in coming months.
2025 is widely regarded as the year of the AI agent. Many startups began shifting their focus to developing intelligent agents last year, and some companies even secured multimillion-yuan orders from enterprise clients for such products.
Still, no consumer-facing AI agent has set the Chinese market alight yet, as Manus is inaccessible to Chinese users, despite attracting intense attention with its initial demo.
Jialong Shi, head of China internet equity research at Nomura, said he thinks the most significant event to watch next for the AI industry will be the launch of a more powerful AI agent.
“It’s already a topic of discussion among investors—everyone is wondering when we’ll see the emergence of a truly powerful AI agent launched by one of China’s AI leaders,” he said.
“Even though none of these companies have spoken publicly about it, I’m sure many of them, including Tencent and ByteDance, are developing their own AI agents internally. So there’s still time. When the moment is right, I believe some of these companies will be able to launch their AI agents, which could hopefully reignite market excitement around AI,” he said.
Whether that reignition moment comes, there is still one big hurdle standing in the way of Chinese AI going global: politics. Weeks after DeepSeek’s breakthrough in January, governments in Asia were already moving to restrict its use due to “security concerns.”
And in May, Semafor reported that the US Treasury Department was reviewing venture capital firm Benchmark’s USD 75 million investment in Manus, which is valued at USD 500 million, to determine whether it complies with 2023 regulations governing investments in Chinese companies.
Manus declined to comment on the matter.
This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.